Our Technology

The best recommender systems for Retails and Brands

Our research focuses on recommender systems, a very dynamic field, easy to understand but complex to solve.

AI Methodology

Step 1

Analyse available datasets

Step 2

Apply our models to generate an initial recommendations set 

Step 3

Use reinforced Machine Learning to refine outputs 

Step 4

Create a virtuous circle of ever more accurate recommendations 

The Data Issue

Data is the #1 AI challenge

Aggregating the right data can be difficult. The data is usually there but it’s often all over the place. And in some cases it’s not there at all.

Most brands and retailers will ultimately have the same AI algorithms. Data will be the key differentiator.

Store Data Issue

Early on we realized very little actionable data were captured from store.

That’s what led us to implement our own simple solution for store data capture. We use tablets or old iOS or Android devices to economically turn stores into data capture powerhouse.

Learn More

Existing Data are central

Existing data already offer a powerful contextual source for recommendations.

Past purchase and web logs already offer enough information for precise recommendations. In fact, that’s exactly what Amazon uses to increase their sales by up to 30% with the now classic ‘Customer who bought this also bought …’

Deep Reinforcement Learning

Our AI learns from customers. It also learns from multiple additional data like location, weather, economic indicators and more. More data combined with ongoing learning means more chances of detecting early weak signals. 

When we can add store data, the recommendations accuracy can increase by up to 18%.